Zhuojin Pipe Fitting Co., LtdVisual Detection of Molten Pool Shape Parameters in Continuous Current TIG Butt Welding
Literature Overview
The paper by Liu Zhiyong et al., published in Welding Technology (Vol. 34, No. 1, 2005, pp. 17–19), presents an early but methodologically significant approach to real-time molten pool monitoring using conventional CCD camera-based vision systems. The research, conducted across industrial and academic institutions in Shandong Province, demonstrates that standard imaging equipment can achieve sufficient resolution for TIG butt weld pool characterization, with custom edge detection algorithms enabling extraction of geometric parameters. This work represents a foundational contribution to in-process monitoring technology that remains relevant for pipe welding quality assurance.
Core Technical Content
Imaging System and Methodology
The authors utilized a standard CCD camera to capture high-resolution images of the TIG butt welding molten pool. The imaging system was calibrated to convert pixel coordinates to actual geometric dimensions, enabling quantitative analysis of pool shape parameters.
Edge Detection Algorithm
A custom image edge detection algorithm was designed specifically for molten pool images, accounting for:
- The intense luminosity of the arc and its halo effect
- The reflective and specular nature of the liquid metal surface
- The dynamic boundary between solidified and molten regions
- Background noise and interference from welding spatter
The algorithm successfully detected the entire perimeter of the continuous current TIG butt weld molten pool, from which geometric parameters such as pool length, width, and perimeter were extracted.
Molten Pool Shape Parameters
The following parameters were characterized through the vision system:
| Parameter | Description | Engineering Significance |
|---|---|---|
| Pool length | Distance from arc center to trailing edge | Indicator of heat input and travel speed |
| Pool width | Maximum transverse dimension | Related to weld bead width and dilution |
| Pool perimeter | Total boundary length | Characterizes pool shape factor |
| Pool area | Enclosed liquid region | Correlates with heat input |
| Aspect ratio | Length-to-width ratio | Indicates penetration tendency |
Engineering Relevance to Pipe Welding
For pipe welding applications—whether for process piping (ASME B31.3), line pipe (API 5L), or structural applications—real-time molten pool monitoring offers several advantages:
- Defect prevention: Abnormal pool geometry (excessively elongated or widened pool) can indicate parameter deviations that lead to undercuts, excessive dilution, or incomplete fusion. Early detection allows corrective action before defects are locked into the solidified weld.
- Process parameter validation: Pool dimensions serve as indirect indicators of heat input, travel speed, and arc stability, providing a non-invasive means of process verification.
- Weld geometry prediction: Molten pool shape parameters correlate with final weld bead dimensions, enabling prediction of weld profile without destructive testing.
- Automation integration: The vision-based approach can be integrated into automated welding systems for feedback control, particularly valuable for pipe welding where joint geometry varies with diameter and thickness.
Key Reflections and Study Insights
While the imaging technology described in this 2005 paper represents the state of the art at that time, the fundamental methodology—optical sensing combined with image processing for in-process monitoring—remains highly relevant. The key insight is that molten pool geometry is a rich information carrier that encapsulates the combined effects of electrical parameters, gas shielding quality, joint fit-up, and material properties. For pipe welding engineers, the practical value lies in translating pool shape observations into actionable process adjustments. The challenge of distinguishing between normal pool variations and defect-indicating anomalies requires both algorithmic sophistication and metallurgical understanding. This paper, though modest in scope, demonstrates the principle that process monitoring need not require exotic instrumentation—careful algorithm design applied to conventional sensors can yield valuable engineering information.